Assuming there ends up being support for multiple signatures within a
single header field to support algorithm dexterity, a less DNS-impacting
option could be to have multiple selector+signature pairs within the field.
This would allow us to keep the single key per record structure of DKIM.

Something like this:

DKIM2-Signature: i=1; d=google.com; s=rsakeysel; bh=RSA_SIG;
s2=ed25519keysel; bh2=ED25519_SIG

On Thu, Jul 10, 2025 at 2:19 PM Wei Chuang <weihaw=
[email protected]> wrote:

>
>
> On Thu, Jul 10, 2025 at 10:52 AM John Levine <[email protected]> wrote:
>
>> It appears that Wei Chuang  <[email protected]> said:
>> >-=-=-=-=-=-
>> >
>> >Hi all,
>> >I'm announcing the Internet-Draft for the "Domain Name specification for
>> >DKIM2".
>> >https://datatracker.ietf.org/doc/draft-chuang-dkim2-dns/02/
>> >
>> >DKIM2 intends to be compatible with the existing DKIM installed base of
>> >keys hence this part of the specification is essentially the same as
>> >RFC6376 with an update for more modern algorithms. ...
>>
>> It mostly looks fine but I have a few questions.
>>
>> It currently only lists k=rsa for RSA signatures.  If we're only going to
>> have one signature scheme, it should be k=ed25519 since those signatures
>> are
>> smaller and faster to compute and verify.
>>
>
> That's an oversight on my part.  I can add that in when the upload service
> is open again.
>
>
>> When we added a new crypto algorithm we realized that you can only have
>> one
>> key per selector.I gather the plan is to allow multiple signatures
>> in the same dkim2-signature header so the key records will need to allow
>> that.
>>
>> One possibility would be to relax the rule about one key TXT record per
>> selector
>> to allow multiple records but the k= values have to be different, e.g.
>> k=ed25519
>> in one and k=postquantum in the other.  Another possibility would be to
>> keep it
>> one record, but make the k= h= p= be comma separated lists rather than
>> single
>> values.
>>
>
> IMO either of these approaches could be sensible.  Happy to document once
> a particular direction is settled upon especially wrt to how the signature
> will be specified.  My guess is that the comma separated will be a little
> bit more straight forward to specify.
> -Wei
>
>
>>
>> R's,
>> John
>>
> _______________________________________________
> Ietf-dkim mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>
_______________________________________________
Ietf-dkim mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to